10 KiB
Table of Contents
- The difficulty of the problem
- Is it really a bug?
- What to send (minimum requirements)
- What to send (additional helpful information)
- Narrowing the problem
- Karma
- Confidentiality of Bug Reports
- Ethics
The difficulty of the problem
This decompiler is a constant work in progress: Python keeps changing, and so does its code generation.
There is no Python decompiler yet that I know about that will decompile everything. Overall, I think this one probably does the best job of any Python decompiler that handles such a wide range of versions.
But at any given time, there are a number of valid Python bytecode
files that I know of that will cause problems. See, for example, the
list in
test/stdlib/runtests.sh.
But I understand: you would the bugs you encounter addressed before all the other known bugs.
From my standpoint, the good thing about the bugs listed in
runtests.sh is that each test case is small and isolated to a single
kind of problem. And I'll tend to fix easier, more isolated cases than
generic "something's wrong" kinds of bugs where I'd have to do a bit
of work to figure out what's up, if not use some sort of mind reading,
make some guesses, and perform some experiments to see if the guesses
are correct. I can't read minds, nor am I into guessing games; I'd
rather devote the effort spent instead towards fixing bugs that are
precisely defined.
And it often turns out that by just fixing the well-defined and prescribed cases, the ill-defined amorphous cases as well will get handled as well.
In sum, you may need to do some work to have the bug you have found handled before the hundreds of other bugs, and other things I could be doing.
No one is getting paid to work to work on this project, let alone the bugs you may have an interest in. If you require decompiling bytecode immediately, consider using a decompilation service, listed further down in this document.
Is it really a bug?
Do you have valid bytecode?
As mentioned in README.rst, this project doesn't handle obfuscated code. See README.rst for suggestions for how to remove some kinds of obfuscation.
Checking if bytecode is valid is pretty simple: disassemble the code.
Python comes with a disassembly module called dis. A prerequisite
module for this package, xdis has a cross-python version
disassembler called pydisasm.
Semantic equivalence vs. exact source code
Consider how Python compiles something like "(x*y) + 5". Early on Python creates an "abstract syntax tree" (AST) for this. And this is "abstract" in the sense that unimportant, redundant or unnecessary items have been removed. Here, this means that any notion that you wrote "x+y" in parenthesis is lost, since in this context they are unneeded. Also lost is the fact that the multiplication didn't have spaces around it while the addition did. It should not come as a surprise then that the bytecode which is derived from the AST also has no notion of such possible variation. Generally this kind of thing isn't noticed since the Python community has laid out a very rigid set of formatting guidelines; and it has largely beaten the community into compliance.
Almost all versions of Python can perform some sort of code improvement that can't be undone. In earlier versions of Python it is rare; in later Python versions, it is more common.
If the code emitted is semantically equivalent, then this isn't a bug.
For example the code might be
if a:
if b:
x = 1
and we might produce:
if a and b:
x = 1
These are equivalent. Sometimes
else:
if ...
may come out as elif or vice versa.
As mentioned in the README, It is possible that Python changes what you write to be more efficient. For example, for:
if True:
x = 5
Python will generate code like:
x = 5
Even more extreme, if your code is:
if False:
x = 1
y = 2
# ...
Python will eliminate the entire "if" statement.
So just because the text isn't the same, does not necessarily mean there's a bug.
What to send (minimum requirements)
The basic requirement is pretty simple:
- Python bytecode
- Python source text
Please don't put files on download services that one has to register
for or can't get to by issuing a simple curl or wget. If you can't
attach it to the issue, or create a github gist, then the code you are
sending is too large.
Also try to narrow the bug. See below.
What to send (additional helpful information)
Some kind folks also give the invocation they used and the output which usually includes an error message produced. This is helpful. From this, I can figure out what OS you are running this on and what version of uncomplye6 was used. Therefore, if you don't provide the input command and the output from that, please give:
- uncompyle6 version used
- OS that you used this on
- Python interpreter version used
But I don't have the source code!
Sure, I get it. No problem. There is Python assembly code on parse
errors, so simply by hand decompile that. To get a full disassembly,
use pydisasm from the xdis
package. Opcodes are described in the documentation for
the dis module.
But I don't have the source code and am incapable of figuring how to do a hand disassembly!
Well, you could learn. No one is born into this world knowing how to disassemble Python bytecode. And as Richard Feynman once said, "What one fool can learn, so can another."
If this is too difficult, or too time consuming, or not of interest to you, then perhaps what require is a decompilation service. Crazy Compilers offers a byte-code decompiler service for versions of Python up to 2.6. (If there are others around let me know and I'll list them here.)
Narrowing the problem
I don't need or want the entire source code base for the file(s) or module(s) can't be decompiled. I just need those file(s) or module(s). If there are problems in several files, file a bug report for each file.
Python modules can get quite large, and usually decompilation problems occur in a single function or maybe the main-line code but not any of the functions or classes. So please chop down the source code by removing those parts that do to decompile properly.
By doing this, you'll probably have a better sense of what exactly is the problem. Perhaps you can find the boundary of what decompiles, and what doesn't. That is useful. Or maybe the same file will decompile properly on a neighboring version of Python. That is helpful too.
In sum, the more you can isolate or narrow the problem, the more likely the problem will be fixed and fixed sooner.
Karma
I realize that following the instructions given herein puts a bit of burden on the bug reporter. In my opinion, this is justified as attempts to balance somewhat the burden and effort needed to fix the bug and the attempts to balance number of would-be bug reporters with the number of bug fixers. Better bug reporters are more likely to move in the category of bug fixers.
The barrier to reporting a big is pretty small: all you really need is a github account, and the ability to type something after clicking some buttons. So the reality is that many people just don't bother to read these instructions, let alone follow it to any simulacrum.
And the reality is also that bugs sometimes get fixed even though these instructions are not followed.
So one factors I may take into consideration is the bug reporter's karma.
- Have you demonstrably contributed to open source? I may look at your github profile to see what contributions you have made, how popular those contributions are, or how popular you are.
- How appreciative are you? Have you starred this project that you are seeking help from? Have you starred any github project? And the above two kind of feed into ...
- Attitude. Some people feel that they are doing me and the world a great favor by just pointing out that there is a problem whose solution would greatly benefit them. Perhaps this is why they feel that instructions are not to be followed by them, nor any need for showing evidence gratitude when help is offered them.
Confidentiality of Bug Reports
When you report a bug, you are giving up confidentiality to the source code and the byte code. However, I would imagine that if you have narrowed the problem sufficiently, confidentiality of the little that remains would not be an issue.
However feel free to remove any comments, and modify variable names or constants in the source code.
Ethics
I do not condone using this program for unethical or illegal purposes. More detestable, at least to me, is asking for help to assist you in something that might not legitimate.
Don't use the issue tracker for such solicitations. To try to stave off illegitimate behavior, you should note that the issue tracker, the code, and bugs mentioned in that are in the open: there is no confidentiality. You may be asked about the authorship or claimed ownership of the bytecode. If I think something is not quite right, I may label the issue questionable which may make the it easier those who are looking for illegal activity.